genetic structure of the annual weed senecio vulgaris in relation to

Molecular Ecology (2001)

10

, 17–28

© 2001 Blackwell Science Ltd

Blackwell Science, Ltd

Genetic structure of the annual weed

Senecio vulgaris

in relation to habitat type and population size

H . MÜLLER-SCHÄRER* and M. F ISCHER†*

Départment de Biologie/Ecologie, Université de Friboug, Pérolles, CH-1700 Fribourg,

†

Institut für Umweltwissenschaften, Universität Zürich, Winterthurerstrasse 190, CH-8057 Zürich, Switzerland

Abstract

Throughout the world, the highly selfing annual common groundsel,

Senecio vulgaris

(Asteraceae) is a common weed. Recently, it has also colonized ecological compensationareas in agro-ecosystems. We investigated the genetic structure of

S. vulgaris

using randomamplified polymorphic DNA (RAPD) profiles of 80 plants from nine populations representingthree habitat types in two regions in Switzerland. RAPD variation among regions (19.8%),among populations within regions (19.2%) and within populations (61.1%) was highlysignificant (

amova

;

P

< 0.001). Gene flow estimated from the observed differentiation amongpopulations (

Φ

ST

= 0.382) was low (assuming Wright’s island model,

N

e

m

= 0.404). Geneticdistances between pairs of populations were significantly correlated with geographicaldistances (Mantel test;

r

= 0.37,

P

< 0.03). Molecular variance obtained with

amova

waslowest in the small populations in compensation areas (1.13), intermediate in vineyardpopulations (2.49), all located in northern Switzerland and highest in the larger vegetablefield populations from western Switzerland (3.41;

P

< 0.05). Overall, there was a positivecorrelation of molecular variance and population size (

P

< 0.05), as expected under geneticdrift. However, molecular variance was negatively correlated with population size amongpopulations in ecological compensation areas, suggesting that selection was also important.We also applied triazine herbicide to leaves of three offspring of each of the 80 plants. Plantsfrom populations of compensation areas showed higher mean levels and reduced variationin the resistance to triazine herbicide than plants from vineyards and vegetable fields.This suggests that compensation areas were colonized from adjacent corn fields, in whichthere has been selection for herbicide resistance. We discuss the implications of our resultsfor the biological control of

S. vulgaris.

Keywords

: biological control, genetic variation, population genetic structure, population size,RAPD–PCR, weed

Received 23 May 2000; revision received 7 August 2000; accepted 22 August 2000

Introduction

Most weed species grow over a broad range ofenvironmental conditions and exhibit large variation insize and morphology (Barrett 1982). This variation mayeither be a consequence of phenotypic plasticity or resultfrom genetic differentiation within and between populations(Bradshaw 1965; Schlichting 1986; Barrett 1988; Schmid1992; Hermanutz & Weaver 1996). Genetic variation isimportant as a prerequisite for further evolution, for

successful establishment after seed dispersal and asa major determinant of plant response to pathogens(Silvertown & Lovett Doust 1993). Studies of variationin flower morphology, isozymes (Warwick 1990), DNAmarkers (Cavan & Moss 1997; Dietz

et al

. 1999), physiology(Darmency & Aujas 1986), effects of herbicides (Gasquez1991), or bioassays with pathogens (Burdon

et al

. 1983)show that most weeds are genetically variable, with theexception of those which reproduce exclusively asexually(Cousens & Croft 2000).

Senecio vulgaris

L. (Asteraceae), common groundsel, is avery short-lived, predominantly selfing annual. While itmost probably originated in southern Europe (Kadereit

Correspondence: H. Müller-Schärer. Fax: + 41 26 30097 40; E-mail:[email protected]

MEC1169.fm 7 Friday, December 8, 2000 3:19 PM

18

H . M Ü L L E R - S C H Ä R E R and M . F I S C H E R

© 2001 Blackwell Science Ltd,

Molecular Ecology

, 10, 17–28

1984), it is now common throughout the world. Becauseof its short generation time, its capacity for prolific seedproduction and rapid germination throughout the year,

S. vulgaris

is a troublesome weed, especially in plantnurseries, orchards, vineyards and fields with frequentcultivation (Holm

et al

. 1997).

Senecio vulgaris

was the firstspecies which developed herbicide resistance and now-adays triazine herbicide-resistant populations are commonand widespread in Europe and North America (Holt &LeBaron 1990). While dunes probably constitute the naturalhabitat of

S. vulgaris

, it is mainly associated with dis-turbed habitats such as ruderal sites (gravel pits, wastegrounds, roadsides) and frequently cultivated crop fields,where it grows in large populations. More recently, smallpopulations of

S. vulgaris

have established in so-calledareas of ecological compensation in agro-ecosystems. InSwitzerland, a large number of compensation strips hasbeen installed during the past 10 years to promote speciesdiversity in arable and grassland systems, such as bysowing wild flower strips (

Buntbrachen

) or by establishingcrop strips without input of fertilizers and herbicides(

Ackerschonstreifen

). Due to its potential as a weed,

S. vulgaris

has never been used in such seed mixtures of wild flowers.Because of the problems associated with intensive use

of herbicides, biological control of weeds is considered avaluable alternative, i.e. the deliberate use of naturalenemies to reduce biomass or population density of a targetweed (Müller-Schärer

et al

. 2000).

Puccinia lagenophorae

(Uredinales, Basidiomycetes), a naturalized rust fungusfrom Australasia, is presently the most common pathogenon

S. vulgari

s in central Europe. In Switzerland, it was foundassociated with

S. vulgaris

in various habitats, includingcrop fields and ecological compensation areas (Müller-Schärer & Wyss 1994). With

S. vulgaris

and

P. lagenophorae

as a model system, a ‘system management approach’ ofbiological weed control is being developed which aims atmanagement of the weed-pathosystem in order to maximizethe natural spread and disease severity of a pathogen(Müller-Schärer & Frantzen 1996; Frantzen & Hatcher1997; Frantzen & Müller-Schärer 1998; Müller-Schärer &Rieger 1998). Knowledge of amount and distribution ofphenotypic and genetic variation in

S. vulgaris

popula-tions is important both to achieve this goal (Frantzen& Müller-Schärer 1998) and for measures to minimizethe development of resistance to the pathogen (Wyss &Müller-Schärer 1999). It is of particular interest whethercompensation area populations are genetically differentfrom crop field populations and which proportion of geneticvariation of

S. vulgaris

is harboured in ecological compensa-tion areas, because genetic variation in compensation areasmay affect the success of biological control of

S. vulgaris

notonly there, but also through gene flow in adjacent crop fields.

Here, we studied genetic variation in plants represent-ing nine populations of

S. vulgaris

from three different

habitat types in Switzerland: ecological compensationareas (

Buntbrachen

and

Ackerschonstreifen

), vineyards andvegetable fields. The last two represent a perennial andan annual cropping system, respectively. In all three habitattypes,

S. vulgaris

is a troublesome weed and biologicalcontrol is a potential control measure. Because

S. vulgaris

is highly selfing and because its populations are separatedby unsuitable habitat such as forests, grasslands andcereal fields, we expect pronounced differentiation amongpopulations. Moreover, we hypothesize that genetic diversitymay differ between populations of

S. vulgaris

in differenthabitat types. On the one hand, low management intensityand low levels of disturbance and herbicide input in areasof ecological compensation enable high plant diversityand high pollinator activity, which may increase the chancefor successful seed-set by

S. vulgaris

and might favouroutcrossing in this predominantly selfing species (Campbell& Abbott 1976), which in turn may increase genetic diversity.On the other hand, populations in ecological compensa-tion areas may constitute sink populations from adjacentcorn fields in which there has been strong selection forherbicide resistance. Occurrence of triazine herbicide resist-ance in populations from ecological compensation areaswould provide support for this hypothesis. Coloniza-tion by a few (herbicide-resistant) founders may lead tolower genetic variability in

S. vulgaris

populations in com-pensation areas compared with populations in crop fields.The level of disturbance by management interventionsis highest and the herbicide use most intensive in thevegetable fields, in which more than one crop is grown peryear. Levels of agricultural interventions are intermediate invineyards and lowest or negligible in the compensation areas.

We studied random amplified polymorphic DNA (RAPD;Williams

et al

. 1990). Once this method is established,RAPD–PCR (polymerase chain reaction) is quick and easy,requires little plant material and offers a high resolution(Steinger

et al

. 1996). We ask how molecular genetic (RAPD)variation is partitioned between two sampling regions,among populations within regions and within populations.Moreover, we estimate gene flow and ask whether geneticand geographical distances are correlated. Then we askwhether observed RAPD patterns reflect effects of selection,or effects of genetic drift. If drift is more important we donot expect genetic differences between populations to berelated to the different habitat types, but we expect lowergenetic variation in smaller populations than in larger ones.In contrast, if selection is more important, genetic differencesbetween populations may be related to the different habitattypes and genetic variation may be lower in larger popu-lations than in smaller ones.

Inevitably, the interpretation of the RAPD structure of

S. vulgaris

is complicated by the fact, that different habitattypes occur in different geographical regions and thatpopulations in different habitat types differ in their sizes.

MEC1169.fm Page 18 Friday, December 8, 2000 3:19 PM

G E N E T I C S T R U C T U R E O F

S E N E C I O V U L G A R I S

19


Molecular Ecology

, 10, 17–28

Therefore, to obtain further information on the importanceof selection, we complement the molecular data with anexperiment on levels and variation in triazine resistance,using offspring of the plants analysed with RAPDs. Therewe ask whether average levels and variation in triazineherbicide resistance of plants of

S. vulgaris

vary betweenhabitat types and populations of different size. Then, wediscuss the genetic structure of

S. vulgaris

populations inthe three habitat types and evaluate possible implicationsfor biological control of

S. vulgaris

with the naturalizedrust fungus

P. lagenophorae.

Materials and methods

Selection and size of study populations and sampling of plant material

Many ecological compensation areas, which have beenestablished in large numbers in the Klettgau area ofnortheast Switzerland in the late 1980s, have been colonizedby

Senecio vulgaris

. In the Klettgau area

S. vulgaris

is also acommon weed in vineyards. As a weed in vegetable fields

S. vulgaris

can frequently be found in the canton of Fribourg,about 130 km southwest of the Klettgau. We studied ninepopulations of

S. vulgaris

, three in the canton of Fribourgand six in the Klettgau (Table 1). The three selectedpopulations in vegetable fields in the canton of Fribourgwere situated within a diameter of 6 km and the sixpopulations in compensation areas and in vineyards inthe Klettgau were randomly dispersed within a diameterof 14 km. Pollinator activity (measured on 23 July 1998as number of potentially pollinating insects flying over

five 4-m lines per population during 2 min per line) wassignificantly higher in compensation areas (19.9 insects/min) than in vineyards (6.4 insects/min; S. Rigamonti,unpublished data).

In early May 1997, we estimated population size of

S. vulgaris

by multiplying average densities with the areaoccupied by

S. vulgaris

. Such areas were either entireecological compensation plots (strips 6–12 m wide) or homo-geneously managed crop fields. To assess average dens-ity at a site, we counted all

S. vulgaris

plants in six 1 m

2

plots regularly placed in a 1-by-100 m transect. In eachpopulation, we collected seeds from 20 plants randomlyselected within the transect (in vegetable fields in fall1996 and in the Klettgau in early May 1997). Thus, plantsamples of all nine populations represent areas of thesame size. During October and November 1997 we grewmaternal seed families from all sampled plants underhomogeneous conditions in a heated greenhouse.

DNA extraction and amplification

For RAPD analysis, we selected up to 10 maternal seedfamilies per population (from some of the populationsless than 10 seed families had germinated). Then, werandomly selected one plant per maternal seed family on26 October 1997. Finally, we took one young leaf from eachof 80 plants, representing 7–10 seed families of each of thenine populations (Table 1). The leaves were rinsed withdistilled water, put into Eppendorf tubes, lyophilized andstored at –18

°

C. DNA was extracted in November 1997using a CTAB procedure (modified from Rogers & Bendich1988; see Steinger

et al

. 1996). Amplification reactions

Table 1 Study populations of Senecio vulgaris

Population SiteLongitude (east)

Latitude (north)

Colonized area (m2)

Population size

n plants examined

n RAPD phenotypes

Molecularvariance

KlettgauCA1 Klettgau 1 8°30’57′′ 47°41′59′′ 500 240 9 3 0.94CA2 Klettgau 2 8°30′54′′ 47°42′01′′ 500 315 7 2 0.80CA3 Rafz/Eglisau 8°30′52′′ 47°35′17′′ 200 68 8 3 1.66VI1 Siblingen 8°31′44′′ 47°42′30′′ 1000 2500 10 5 2.21VI2 Oberhallau 8°28′14′′ 47°42′45′′ 1000 320 10 7 2.62VI3 Hallau 8°27′49′′ 47°42′26′′ 1000 600 9 8 2.63

FribourgVE1 Sugiez 7°06′11′′ 46°57′15′′ 100 468 7 7 3.38VE2 Ried 7°11′03′′ 46°56′59′′ 270 2376 10 7 3.44VE3 Kerzers 7°10′51′′ 46°58′42′′ 205 3690 10 7 3.41

Colonized area, indicates the total area of the population of S. vulgaris; population size, denotes the number of flowering plants in May 1997; n plants examined, denotes the number of plants used for molecular analyses; n RAPD phenotypes, denotes the number of different RAPD phenotypes to which the examined plants belonged; molecular variance (amova sum of squares divided by N-1) is a measure of genetic variability (see Methods); CA, denotes populations in ecological compensation areas; VE, those in vegetable crops; and VI, those in vineyards.


20

H . M Ü L L E R - S C H Ä R E R and M . F I S C H E R


Molecular Ecology

, 10, 17–28

(25

µ

L) contained DyNAzyne™ II reaction buffer [10 m

m

Tris–HCl (pH 8.8), 50 m

m

KCl, 1.5 m

m

MgCl

2

, 0.1% TritonX-100), 100

µ

m

each dATP, dCTP, dGTP, dTTP, 0.2

µ

m

primer (Operon Technologies Inc.), 0.5 units DyNazyme™II polymerase (Finnzymes OY) ] and approximately 20 ngDNA. Reactions were held in polycarbonate microtitreplates and overlaid with one drop of mineral oil. Thesamples were incubated in a DNA thermocycler (MJResearch Inc.; model PTC100–96 V) programmed withthe folllowing parameters: 1 min at 93

°

C followed by 45cycles of 30 s at 92

°

C, 30 s at 36

°

C and 90 s at 72

°

C. Anincubation of 5 min at 72

°

C was included as a final step.Amplified DNA fragments were separated by electrophoresison 1.6% agarose gels and subsequently visualized byethidium bromide staining. Gels were digitized on a UVtable with a video camera-based imaging system. In a firstseries of amplifications 20 10-base primers (Kit A fromOperon) were tested for reproducibility of the amplifiedfragment profile using a minimum of four replicates ofa single DNA extract. The first 10 primers yielding repro-ducible patterns (primers A1, A2, A5, A8, A10, A11, A14,A15, A17 and A18) were selected for the RAPD analysis ofall 80 sampled plants. The presence or absence of bandswas scored for 74 band positions in the range of inter-mediate molecular weight (Stewart & Porter 1995).

Triazine herbicide resistance

On 21 October 1998, one year after the RAPD analyses, 7–10 seeds, resulting from selfing of the 80 plants used forthe molecular analyses, were sown in seed trays and keptin the greenhouse of the Department of Biology of theUniversity of Fribourg. Two weeks later, three seedlingsper maternal plant were planted individually into 9-cmdiameter plastic pots, which were filled with nutrient-amended peat (Floragard TKS2; Floragard). The pots wereplaced randomly on a tray in the greenhouse and kept ata 16-h day (with a light intensity of c. 250

µ

mol m

–2

s

–1

at23

°

C and 60% humidity) and an 8-h night (at 16

°

C and80% humidity). On 17 November 1998, when plants hadgrown on average eight true leaves, they were treatedwith triazine (Gesaprim; 50% atrazine, Novartis Agro AG),using a concentration corresponding to a normal fieldapplication rate of 1.5 kg active ingredient ha

–1

. For thetreatment, plants were arranged on a surface of 0.62 m

2

and sprayed for 31 s with an aerospray [Aero-Spray 2000;Birchmeier & Cie; standard nozzle (0.16) with 1 mL/s at200 kPa and a spray distance of 30 cm]. This correspondsto a herbicide dosage of 500 L/ha. Effects of the herbicidewere assessed 15 days after application. Phytotoxic symptomsof all plants were individually ranked according to theEuropean Weed Research Society (EWRS) classificationscheme for plant tolerance (scaling from 1 to 9; with 1indicating no symptoms/healthy plant and 7–9 representing

heavy damage to death; Anonymous 1992). This measure-ment is a good indicator of the level of triazine resistance(Frey

et al

. 1999). Plants in our study showed phytotoxicsymptoms in the range of 1–8 at 15 days after herbicideapplication, when the experiment was stopped. Accordingto our previous studies (Frey

et al

. 1999), plants with valuesabove 3 will die prior to seed set.

Statistical analysis

Variation in RAPD patterns was analysed with analysisof molecular variance (program

amova

, version 1.55, seeExcoffier

et al

. 1992; Stewart & Excoffier 1996).

amovaanalyses are based on the pairwise squared Euclideandistances between RAPD phenotypes. Because band statescan only take the values 0 and 1, these Euclidean dis-tances equal the number of band states in which pairs ofRAPD phenotypes differ. Because sums of squares in aconventional analysis of variance (anova) can be writtenas sums of squared pairwise differences, amova is closelyrelated to anova. amova allowed us to calculate variancecomponents and their significance levels for variationamong regions, among populations within regions and withinpopulations. We also used amova to test for differencesbetween vineyard and compensation area populationswithin the Klettgau region. Because significance tests inamova are based on permutation procedures, they areessentially distribution free (Excoffier et al. 1992). Theprogram also extracts analogues of F-statistics, so-calledΦ-statistics. Thus, pairwise genetic distances (ΦST) amongthe nine populations and their levels of significance werealso obtained from the amova. Under the assump-tions of Wright’s island model, gene flow (Nem) can beapproximated as Nem = [1/(FST – 1) ]/4 (Wright 1951; butsee Whitlock & McCauley 1999). Homogeneity of molecularvariance in pairs of populations was tested with Bartlett’stest, which is also implemented in the program amova 1.55.

We used a Mantel permutation test (Mantel 1967) to testwhether genetic distances between pairs of populationswere significantly correlated with corresponding geograph-ical distances [program mantel of the ‘R’ package (release3.0) for multivariate analysis; Legendre & Vaudor 1991].

We analysed molecular variance with anova, fitting the(log-transformed) covariate population size, the factorhabitat type and their interaction. To consider covariateand factor before the highly significant two–way interactionwe used sequential sums-of-squares (Payne et al. 1993).The levels of significance of habitat type and populationsize did not depend on their sequence in the model.

To analyse phytotoxicity levels after herbicide applica-tion, we used a hierarchical anova with populations nestedwithin habitat types and seed families nested within popu-lations. Because at the time of herbicide application thenumber of leaves did not vary among plants from the


G E N E T I C S T R U C T U R E O F S E N E C I O V U L G A R I S 21

© 2001 Blackwell Science Ltd, Molecular Ecology, 10, 17–28

three habitat types, developmental stage was not con-sidered in this analysis. To quantify variation in the phy-totoxicity levels among plants, we used the Berger-ParkerDominance index, which indicates the proportion of themost frequent phytotoxicity phenotype in a population(Southwood 1978). Higher values of this index indicatelower levels of variation.

Results

Partitioning of RAPD variation and gene flow

The final presence–absence matrix contained scores of80 individuals at each of the 74 bands (primer A1, eightbands; A2, nine bands; A5, six bands; A8, three bands; A10,nine bands; A11, ten bands; A14, three bands; A15, eightbands; A17, 11 bands; A18, seven bands). The 80 plants ofSenecio vulgaris belonged to 42 different RAPD phenotypes.Twenty-six of the 74 scored bands were polymorphic (35%).None of the nine populations was monomorphic. For each

population sample of 6–10 plants, we detected betweentwo and eight different RAPD phenotypes (Table 1). Thepair of RAPD phenotypes with the largest Euclideandistance differed in 15 of the 26 polymorphic bands.

Variation in RAPD banding pattern among the tworegions, among populations within regions and withinpopulations was highly significant (P < 0.001) at each level,indicating genetic divergence both among regions andamong populations (Table 2). ΦST, the correlation amongrandom RAPD phenotypes within populations relative tothe correlation of random pairs drawn from the wholesample was 0.382. ΦCT, the correlation among random pheno-types within regions relative to the correlation of randompairs drawn from the whole sample was 0.210 and ΦSC, thecorrelation of random phenotypes within populations,relative to that of random pairs drawn from the region,was 0.219. Gene flow, i.e. the average number of individualsexchanged among populations per generation (Nem), waslow (0.404). To estimate gene flow among populations withinregions, we used ΦSC instead of ΦST and obtained Nem as0.892. Within the Klettgau region, there was a significantdifference between S. vulgaris populations in vineyardsand in ecological compensation areas (amova, 21% of thevariation between habitat types; P < 0.001.

Twenty-eight of the 36 pairwise genetic distances (ΦST)between populations were significant (Table 3). The threepopulations from ecological compensation areas did notsignificantly differ from each other, although one wasmore than 12 km away from the others.

Correlation between genetic and geographical distances between pairs of populations

Pairwise genetic distances were positively correlated withgeographical distances (nine populations; r = 0.366; P = 0.028;Mantel-test). When we excluded the three populationsfrom vegetable fields in the canton of Fribourg and only

Table 2 Summary of hierarchical analysis of molecular variance(amova)

Variance components

Level of variation d.f. Absolute % P

Among regions 1 0.786 19.77 <0.001Among populations 7 0.762 19.18 <0.001Within populations 71 2.428 61.05 <0.001Total 79

Plants represented nine populations of Senecio vulgaris from two regions, canton of Friboug (three populations) and Klettgau (six populations). The analysis is based on RAPD phenotypes consisting of scores at 26 polymorphic band positions. Levels of significance are based on 1000 iteration steps.

Population

CA1 CA2 CA3 VI1 VI2 VI3 VE1 VE2

CA1 —CA2 0.05 —CA3 0.01 0.07 —VI1 0.02 0.03 0.03 —VI2 0.41*** 0.40*** 0.39*** 0.21* —VI3 0.54*** 0.50*** 0.48*** 0.36*** 0.23*** —VE1 0.41*** 0.36*** 0.32*** 0.21* 0.15* 0.21* —VE2 0.50*** 0.49*** 0.44*** 0.36*** 0.18*** 0.34*** 0.06 —VE3 0.50*** 0.49*** 0.43*** 0.32*** 0.20*** 0.36*** 0.12* 0.12

*P < 0.05; ***P < 0.001. P-values indicate the probability that a random genetic distance ΦST is larger than the observed distance and are based on 1000 iterations. Populations are denoted as in Table 1.

Table 3 Pairwise genetic distances (ΦST)among nine populations of Senecio vulgaris


22 H . M Ü L L E R - S C H Ä R E R and M . F I S C H E R


analysed distances among the six Klettgau populations,the resulting positive correlation of genetic and geographicaldistances was weak and not statistically significant (sixpopulations; r = 0.0075; P = 0.40; Mantel-test).

Molecular variation

Molecular variance per population sample (which repres-ented similar areas in all populations) differed signific-antly among populations of different regions, of differenthabitat types and among populations within habitattypes (P < 0.001; Bartlett test). Twenty-eight of 36 pairwiseBartlett tests indicated significant differences in populationhomogeneity of molecular variation (Table 4).

Molecular variance was highest in populations in vege-table fields (3.41), intermediate in vineyard populations(2.49) and lowest in populations in ecological compensa-tion areas (1.13; anova and Scheffé F-test; P < 0.001 for allcomparisons). Within habitat types, molecular variancewas negatively correlated with population size (anova,P < 0.05, Table 5). This was pronounced in the ecologicalcompensation areas, but not in the crop populations (asindicated by the habitat type-by–population size interac-

tion). Overall, the correlation between population size andmolecular variance was significantly positive (Spearman’srank-correlation coefficient rs = 0.72, P < 0.05; Fig. 1a).

Mean and variation in the level of herbicide resistance

Plants from populations of different habitat types differedin their level of phytotoxicity 15 days after application ofatrazine to their leaves (anova; P < 0.03). Mean levels ofphytotoxicity were lowest (i.e. the level of resistance washighest) in plants from populations in ecological com-pensation areas (mean score 2.9). Levels of phytotoxicitywere intermediate in plants from vineyard populations(mean score 4.3) and they were highest in plants fromvegetable field populations (mean score 6.1). While themean level observed for plants from compensation areasindicates resistance, mean levels of plants from vegetablefields and vineyards are so high, that most plants wouldhave died before reproduction. Levels of phytotoxicitydiffered significantly among seed families within popu-lations (P < 0.01), suggesting genetic variation in resistancealso within populations (Table 6). Within habitat types,phytotoxicity levels were independent of population size.However, overall phytotoxicity levels were higher in largerpopulations (Spearman’s rank-correlation coefficient rs = 0.56,P < 0.04) and in populations with higher molecular vari-ance (rs = 0.72, P < 0.01; Fig. 1b).

The Berger–Parker Dominance index, which indicatesthe proportion of the most frequent phytotoxicity phenotype(i.e. higher values of this index indicate lower variation),tended to be higher in populations in compensation areas(0.65) than in those in vineyards (0.47) or in vegetablefields (0.35). Within habitat types, this index was inde-pendent of population size. However, overall the correla-tions of the Berger–Parker Dominance index with populationsize (Spearman’s rank-correlation coefficient rs = –0.74,P < 0.03) and with molecular variance (rs = –0.44, P < 0.001;

Population

CA1 CA2 CA3 VI1 VI2 VI3 VE1 VE2

CA1 —CA2 1.27 —CA3 1.55 2.02 —VI1 2.46 2.48 1.34 —VI2 7.85*** 6.36*** 5.56*** 3.19* —VI3 10.63*** 8.01*** 7.19*** 5.40*** 3.43*** —VE1 7.47*** 5.88*** 4.34** 3.00* 2.20* 2.77** —VE2 10.69*** 8.61*** 7.09*** 5.71*** 2.89*** 5.11*** 1.35 —VE3 10.61*** 8.62*** 6.87*** 5.16*** 3.18*** 5.48*** 1.86 2.09*

*P < 0.05; **P < 0.01; ***P < 0.001. P-values indicate the probability that a random B is larger than the observed B and are based on 1000 iterations. Populations are denoted as in Table 1.

Table 4 Pairwise tests of heteroscedasticityof molecular variance among nine populationsof Senecio vulgaris. Bartlett’s B is given foreach pair of populations

Table 5 Summary of analysis of variance of the relationship ofhabitat type and population size (after log-transformation) withmolecular variance (a measure of RAPD variability within popula-tions, see Methods) of nine Swiss populations of Senecio vulgaris

Source of variation d.f.Sum of squares F P

Habitat type 2 7.89 954.4 0.0001Log (pop. size)* 1 0.237 57.3 0.0047Habitat type-by-log (pop. size) 2 0.292 35.3 0.0081Error 3 0.0124

*pop. size, population size.




Fig. 1c) were negative, indicating smaller variation in theresponse to triazine herbicide among plants from smallerpopulations with smaller molecular variance.

Discussion

Analysis of genetic variation with RAPD

Since the development of appropriate statistical procedures(see Excoffier et al. 1992; Lynch & Milligan 1994; Stewart& Excoffier 1996) RAPD (Williams et al. 1990) has beenwidely used for the analysis of population genetic structures(Bussell 1999). This is of advantage because DNA poly-morphisms, such as those detected by RAPD, are assumedto be less biased estimators of genetic variation thanvariation at the level of gene products such as allozymes(Stewart & Excoffier 1996). In our study, RAPD proved tobe a high-resolution method for the detection of geneticvariation in Senecio vulgaris and indicated pronouncedgenetic differentiation among populations. However, theestimation of population genetic parameters from RAPDdata is under debate because the inference of FST fromgenerally dominant RAPD data is based on the twoassumptions of homologous null bands and of Hardy–Weinberg equilibrium (Lynch & Milligan 1994; Ayres &Ryan 1999). The high levels of significance in the amova(Table 2) suggest that our results are robust to deviationsfrom the two assumptions. Nevertheless, deviations fromHardy–Weinberg equilibrium due to the high degreesof selfing in S. vulgaris (Campbell & Abbott 1976) mayinflate RAPD-based FST estimates. However, our estimateof FST for S. vulgaris was smaller than the average observedfor highly selfing annuals (Hamrick & Godt 1990; Bussell1999; as discussed below).

In our study we used 74 markers of which 26 were poly-morphic (RAPDs). Because the variance of populationgenetic estimates hardly decreases any more in this rangeof marker numbers (Aagard et al. 1998), we consider thenumber of markers in our population genetic study assufficient. Because RAPD-based estimates of differentiationmight be biased towards higher values compared with

Table 6 Summary of hierarchical analysis of variance of phytotoxicsymptoms (level of resistance against triazine herbicides) of plantsfrom nine Swiss populations of Senecio vulgaris from differenthabitat types assessed 15 days after herbicide application

Level of variation d.f.Sum of squares F P

Habitat type 2 65.56 5.88 0.039Populations within habitat types 6 33.47 2.01 0.076Seed families within populations 71 197.31 1.78 0.002Error 160 250.00

Fig. 1 The relationship between molecular variance (a measureof RAPD variability within populations of Senecio vulgaris, whichis based on the same sampling area in all nine study populations,see Methods) and (a) population size (number of flowering plantsper population in May 1997); (b) level of mean herbicide resistance(where higher values of the phytotoxic symptoms indicate lowerresistance); and (c) Berger–Parker Dominance index (which indicatesthe proportion of the most frequent phenotype in terms of herbicideresistance, see Methods; higher values of this index indicate lowervariation). Measurements of phytotoxicity of triazine herbicide weremade on selfed offspring of the 80 plants used for RAPD–PCRanalyses. Population symbols are labelled according to habitattypes: CA denotes populations in ecological compensation areas;VE, those in vegetable crops; and VI, those in vineyards.




estimates based on other markers, it was suggested to use 100plants per population and to restrict data sets accordingto the ‘3/N-criterion’, i.e. to exclude markers with verylow frequencies of null alleles (Lynch & Milligan 1994; seeIsabel et al. 1999 for an application). However, genetic dis-tances obtained from a restricted data set were consistentwith those obtained from the complete set in a studyon 11 populations of Hippophae rhamnoides represented byabout 10 plants each (Bartish et al. 1999). In addition, theseauthors conclude that the 3/N-criterion may only beappropriate in analyses with larger population sampleswhile it may even underestimate genetic differentiationamong populations in small samples of closely relatedpopulations (Bartish et al. 1999).

Differentiation among populations

Senecio vulgaris is a short-lived annual and reported tobe predominantly inbreeding (Campbell & Abbott 1976).Hence, genetic differentiation among populations may beexpected (Hamrick et al. 1991; Hamrick & Godt 1996).Indeed, we detected 39% of RAPD variation among popu-lations (Table 2, pooled over both regions). In their reviewcomprising more than 400 plant species Hamrick & Godt(1990) used GST values to indicate the proportion of isozymediversity residing among populations. They report anaverage GST of 22% for perennial herbs compared with 36%for annuals and an average GST of 20% for animal-pollinatedoutcrossers compared with 51% for selfers. To date, RAPD-based GST values are available for 35 plant species, withaverages of 19.3% for 29 outbreeding species and 62.5%for six inbreeding species (Bussell 1999). Compared withthese values the populations of S. vulgaris appear to be lessdifferentiated than expected for a highly selfing annual.This may suggest that the degree of outcrossing in S. vulgarisis higher than previously assumed.

Pronounced genetic differentiation observed amongpopulations of S. vulgaris suggests low gene flow andhigh degrees of population isolation. Indeed, cereal fields,grasslands and forests, which separate populations ofS. vulgaris, constitute barriers against dispersal of seed andpollen.

Genetic and geographical interpopulation distanceswere significantly positively correlated in some plant spe-cies (Godt & Hamrick 1993; Godt et al. 1995; Berge et al.1998; Ayres & Ryan 1999), but not in others (Berge et al.1998; Fischer & Matthies 1998; Fischer et al. 2000; Schmidt& Jensen 2000). Close correspondence of geographicaland genetic distances is only likely if there is at least someoutcrossing in a species, if gene flow is a simple functionof geographical distance and if such an effect of gene flowis not compensated by genetic drift or selection. In ourstudy, genetic and geographical distances were signific-antly correlated at the regional scale including all study

populations of S. vulgaris. Indeed, the large geographicaldistance between the populations in the canton of Fribourgand those in Klettgau may well explain the observedstrong genetic differentiation among these populations(Tables 2 and 3) and the correspondingly low estimate ofgene flow. At the more local scale, among Klettgau popu-lations only, the positive relationship between geographicaland genetic distances was weak and not statisticallysignificant. Populations CA1, CA2 and VI1 (see Table 1),which were located less than 2 km from each other, werenot genetically distinct (Table 3). This might reflect highgene flow over distances of a few kilometres (but see pre-ceding paragraph). Moreover, it could be due to reducedstatistical power. On the other hand, the significant geneticdifferentiation between the interspersed populations invineyards and those in ecological compensation areas inthe Klettgau region suggests that the observed patternof differentiation among populations was at least in partalso due to selection in different habitats. Similarly, selec-tion may have contributed to the differences between thepopulations in the Klettgau region and those in vegetablefields in the canton of Fribourg.

Genetic variation in relation to habitat type and population size

Because the three compensation areas in our study havebeen established only recently, between 1992 and 1996,founder effects are a likely explanation for their smallgenetic variability. In contrast, recruitment of populationsof S. vulgaris from a persistent seed bank, which hasestablished before sites were transformed from crop fieldsinto ecological compensation areas, is unlikely becausemost seedlings of S. vulgaris emerge in the first yearand few seeds retain their viability for up to five years(Roberts 1964; Roberts & Feast 1972).

In corn fields single herbicide types such as triazinesare repeatedly applied over decades, whereas S. vulgarispopulations in vegetable crops are subjected to mixedherbicide and cultivation strategies which makes thedevelopment of one particular form of herbicide resist-ance and the associated reduction of genetic variation inresistance less likely (Moodie et al. 1997). It is well estab-lished that S. vulgaris populations in corn fields are highlytriazine-resistant (Ammon & Beuret 1984; Holt & LeBaron1990; Kees 1991). Nevertheless, it would be interesting tocompare means and variations of the level of triazineresistance of S. vulgaris in Swiss corn fields with those inthe habitat types of our study.

In our study, S. vulgaris populations more resistantto triazine herbicides (i.e. populations in compensationareas adjacent to corn fields) were found to be less genetic-ally polymorphic than susceptible populations (in vine-yards and vegetable fields), indicated by smaller molecular




variance and by reduced variation in the response tothe triazine herbicide. This corresponds well with sim-ilar results in other species (Warwick & Black 1993).Levels of isozyme polymorphism in garden populationsof Chenopodium album were higher than in field popula-tions, which was attributed to the repeated use of triazineherbicides in the field populations (Al Mouemar & Gasquez1983). In a similar study, DNA fingerprinting showed thatherbicide-resistant populations of the primarily selfingAvena fatua and A. ludoviciana were genetically less diversethan those of sensitive plants, whereas resistant and sus-ceptible patches of the outbreeding Alopecurus myosuroideswere equally diverse (Cavan & Moss 1997). Low levelsof genetic variability in triazine-resistant populations ofS. vulgaris may well have been caused by bottleneck effectswhich were especially severe because of the low frequencyof the resistance alleles and because the predominantlyselfing mating system may hamper a quick recovery fromreduced genetic variation (Moodie et al. 1997). In principle,conclusions drawn from variations in selectively neutralmarkers, such as isozymes or DNA polymorphisms, onquantitative genetic variation in fitness relevant traitsmust be considered with caution. However, in our studymolecular variance of populations was closely related withtheir variability in response to triazine herbicide (Fig. 1c).We consider severe founder effects in S. vulgaris in the com-pensation areas of our study as very likely, because compensa-tion areas were situated next to corn fields regularly treatedwith triazine herbicide. Colonization by resistant S. vulgarisgenotypes from such fields may well have caused reducedgenetic variation of populations in compensation areas,despite the higher activity of potential pollinators reportedfrom these diverse and flower-rich habitats.

In S. vulgaris populations in crop fields, especially invegetable fields with their high level of disturbance andherbicide input, we observed the highest levels of perarea genetic variation. In vineyards, with little herbicideinput and an intermediate level of disturbance throughrotational cultivation, both population size and geneticvariation were found to be intermediate compared topopulations from vegetable fields and compensationareas. Comparably high RAPD variation found in untreatedpopulations of Sinapis arvensis and in populations fromcrop fields with rotational cropping and herbicide usecould be explained if seedlings may escape herbicidetreatment, show natural resistance, or germinate fromthe seed bank at a time when herbicide pressure is low(Moodie et al. 1997). An alternative explanation could bedifferent patterns of genetic variation between selectivelyneutral and fitness-related markers. With regard to thepresent study, limited seed dispersal or pollen transferalso might have taken place into the vegetable field popu-lations from S. vulgaris populations of adjacent crops witha different management regime, or from neighbouring

ruderal sites or fallow land, where S. vulgaris often growsat high densities and over several years. Such gene flowthrough seed dispersal might have contributed to therelatively high genetic variation found in vegetable fieldpopulations.

On the other hand, at least in part, the high geneticvariability of populations of S. vulgaris in crop fields maybe due to their large size, which reduces the effects ofgenetic drift. Overall, the correlation between moleculargenetic variation and population size in S. vulgaris waspositive (Fig. 1a). Similarly, reduced genetic variation insmaller populations has been reported for 11 of 16 plantspecies mentioned in a review (Frankham 1996) and in sixof seven species studied since then (Berge et al. 1998;Fischer & Matthies 1998; Persson et al. 1998; Prober et al.1998; Fischer et al. 2000). These relationships were inter-preted as results of genetic drift (Barrett & Kohn 1991;Ellstrand & Elam 1993; Young et al. 1998).

In contrast, we found molecular variance to decreasewith increasing population size in the compensation areasof our study, whereas no such correlation was observedin crop fields (Fig. 1a). This suggests that effects of selec-tion are also important during and after the foundation ofpopulations in compensation areas. Such effects, whicharise if some founding genotypes leave more descendantsthan others, may possibly have been promoted by thelack of disturbance in compensation areas.

Implications for weed–pathogen interactions and biocontrol

Low levels or absence of genetic variation in responseto disease often lead to high levels of disease incidence(Schmid 1994) and severity in crops, which may result indevastating epidemics (Finckh & Wolfe 1997). In contrast,many natural plant populations are genetically diverseand polymorphic for disease resistance (Dinoor & Eshed1984; Burdon 1987; Harry & Clarke 1987; Fritz & Simms1992; Cousens & Croft 2000).

In our study of RAPD and of resistance to triazineherbicide we found high levels of genetic variation oflarge S. vulgaris populations in crop fields and low levelsin smaller sink populations in ecological compensationareas. As far as RAPD variation and variation in responseto the triazine herbicide may also reflect the genetic variationin response to a biocontrol agent, our study suggests thatthe chance of successful biological control of S. vulgarisusing Puccina lagenophorae will at present not be impairedin ecological compensation areas as compared to the studiedcrop habitats. Nevertheless, it might be worthwhile torepeat this study of genetic variation in S. vulgaris popu-lations at a later date, when the compensation areas areolder than only a few years. The observed high densitiesof S. vulgaris populations in the two studied crop habitat




types could favour biocontrol and further build-up thespeed of an epidemic. However, observed high levels ofgenetic variability of S. vulgaris populations in these habitatsmight counter this effect. To test such hypotheses, and tostudy the effects of selection due to P. lagenophorae, com-plementary studies on the dynamics of the S. vulgaris andP. lagenophorae populations, on the genetic compositionof fungal rust populations in the studied habitats andon genetic variation in the response of S. vulgaris to thepathogen are presently underway.

Conclusion

The interpretation of the RAPD structure of Senecio vulgariswas complicated by the facts that different types of habitatoccur in different geographical regions and that populationsrepresenting different habitat types differ in their sizes.For many species, such a situation may be the rule ratherthan the exception. Complementing the RAPD data withquantitative data on resistance phenotypes allowed us toconclude that the low genetic variability and small sizeof the studied compensation area populations are mostprobably due to founder effects by herbicide-resistantgenotypes. In this way, our study exemplifies the usefulnessof combining molecular data with quantitative data.

Acknowledgements

Many thanks to Serena Rigamonti for providing data on phytotoxi-city, to Kirsten Leiss for collecting seed material in the canton ofFribourg, to Marius Käser for RAPD analyses and to KirstenLeiss, Jos Frantzen, Thomas Steinger and Mark van Kleunenfor comments on earlier drafts. We acknowledge financial sup-port by the Swiss Priority Programme Environment/Biodiversity(NF 5001–50250/1.97) and the Swiss National Science Foundation(NF 31–46821.96).

References

Ammon HU, Beuret E (1984) Verbreitung Triazin-resistenterUnkräuter in der Schweiz und bisherige Bekämpfungserfahrungen.Zeitschrift für Pflanzenkrankheiten und Pflanzenschutz, Sonderheft, X,183–191.

Aagard JE, Krutovskii KV, Strauss SH (1998) RAPDs and allozymesexhibit similar levels of diversity and differentiation amongpopulations and races of Douglas fir. Heredity, 81, 69–78.

Al Mouemar A, Gasquez J (1983) Environmental conditions andisozyme polymorphism in Chenopodium album L. Weed Research,23, 141–149.

Anonymous (1992) Manual for Field Trials in Plant Protection. 3rdedn. CIBA-GEIGY Plant Protection, Basel, Switzerland.

Ayres DR, Ryan FJ (1999) Genetic diversity and structure of thenarrow endemic Wyethia reticulata and its congener W. bolanderi(Asteraceae) using RAPD and allozyme techniques. AmericanJournal of Botany, 86, 344–353.

Barrett SCH (1982) Genetic variation in weeds. In: Biological Controlof Weeds with Plant Pathogens (eds Charudattan R, Walker HL),pp. 73–98. John Wiley & Sons Inc., New York.

Barrett SCH (1988) Genetics and evolution of agricultural weeds.In: Weed Management in Agro-Ecosystems: Ecological Approaches(eds Altieri MA, Liebman M), pp. 55–76. CRC Press, Boca Raton.

Barrett SCH, Kohn JR (1991) Genetic and evolutionary con-sequences of small population size in plants: implications forconservation. In: Genetics and Conservtion of Rare Plants (edsFalk DA, Holsinger KE), pp. 3–30. Oxford University Press,New York.

Bartish IV, Jeppsson N, Nybom H (1999) Population genetic struc-ture in the dioecious pioneer plant species Hippophae rhamnoidesinvestigated by random amplified polymorphic DNA (RAPD)markers. Molecular Ecology, 8, 791–802.

Berge G, Nordal I, Hestmark G (1998) The effect of breedingsystems and pollination vectors on the genetic variation ofsmall plant populations within an agricultural landscape. Oikos,81, 17–29.

Bradshaw AD (1965) Evolutionary significance of phenotypicplasticity in plants. Advances in Genetics, 13, 115–155.

Burdon JJ (1987). Diseases and Plant Population Biology. CambridgeUniversity Press, Cambridge.

Burdon JJ, Oates JD, Marshall DR (1983) Interactions betweenAvena and Puccinia species I. The wild hosts: Avena barbata Pottex Link, A. fatua L. & A. ludoviciana Durieu. Journal of AppliedEcology, 20, 571–584.

Bussell JD (1999) The distribution of random amplified poly-morphic DNA (RAPD) diversity amongst populations of Isotomapetraea (Lobeliaceae). Molecular Ecology, 8, 775–789.

Campbell JM, Abbott RJ (1976) Variability of outcrossing frequencyin Senecio vulgaris L. Heredity, 36, 267–274.

Cavan G, Moss SR (1997) Herbicide resistance and gene flow inblack-grass (Alopecurus myosuroides) and wild oats (Avena spp.).Proceedings of the 1997. Brighton Crop Protection Conference — Weeds,pp. 305–310. The British Crop Protection Council, Farnham,UK.

Cousens R, Croft AM (2000) Weed populations and pathogens.Weed Research, 40, 63–82.

Darmency H, Aujas C (1986) Polymorphism for vernalizationrequirements in a population of Avena fatua. Canadian Journal ofBotany, 64, 730–733.

Dietz HJ, Fischer M, Schmid B (1999) The invasion history of a 9-year-old population of Bunias orientalis. Oecologia, 120, 225–234.

Dinoor A, Eshed N (1984) The role and importance of pathogensin natural plant communities. Annual Review of Phytopathology,22, 443–466.

Ellstrand NC, Elam DR (1993) Population genetic consequencesof small population size — implications for plant conservation.Annual Review of Ecology and Systematics, 24, 217–243.

Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecularvariance inferred from metric distances among DNA haplotypes:application to human mitochondrial DNA restriction data.Genetics, 131, 479–491.

Finckh MR, Wolfe MS (1997) The use of biodiversity to restrict plantdisease and some consequences for farmers and society. In:Ecology in Agriculture (ed. Jackson LE), pp. 203–237. AcademicPress, New York.

Fischer M, Husi R, Prati D et al. (2000) RAPD variation amongand within small and large populations of the rare clonal plantRanunculus reptans (Ranunculaceae). American Journal of Botany,87, 1128–1137.

Fischer M, Matthies D (1998) RAPD variation in relation topopulation size and plant fitness in the rare Gentianella germanica(Gentianaceae). American Journal of Botany, 85, 811–819.




Frankham R (1996) Relationship of genetic variation to populationsize in wildlife. Conservation Biology, 66, 1500–1508.

Frantzen J, Hatcher PE (1997) A fresh view on the control of theannual plant Senecio vulgaris. Integrated Pest Management Reviews,2, 77–85.

Frantzen J, Müller-Schärer H (1998) A theory relating focal epidemicsto crop–weed interactions. Phytopathology, 88, 180–184.

Frey JE, Müller-Schärer H, Frey B, Frei D (1999) Complex relationbetween triazine-susceptible phenotype and genotype in theweed Senecio vulgaris may be caused by chloroplast DNA poly-morphism. Theoretical and Applied Genetics, 99, 578–586.

Fritz RS, Simms EL (1992) Ecological Genetics of Plant–PhytophageInteractions. In: Plant Resistance to Herbivores and Pathogens. Ecology,Evolution and Genetics (eds Fritz RS, Simms EL), pp. 1–9. TheUniversity of Chicago Press, Chicago.

Gasquez J (1991) Mutation for triazine resistance within susceptiblepopulations of Chenopodium album L. Proceedings of the 19th LongAshton Symposium on Herbicide Resistance in Weeds and Crops,pp. 103–115. Butterworth-Heinemann Ltd, Oxford, UK.

Godt MJW, Hamrick JL (1993) Genetic diversity and populationstructure in Tradescantia hirsuticaulis (Commelinaceae). AmericanJournal of Botany, 80, 959–966.

Godt MJW, Hamrick JL, Bratton S (1995) Genetic diversity ina threatened wetland species, Helonias bullata (Liliaceae). Con-servation Biology, 9, 596–604.

Hamrick JL, Godt JW (1990) Allozyme diversity in plant species.In: Plant Population Genetics, Breeding and Genetic Resources (edsBrown AHD, Clegg MT, Kahler AL, Weir BS), pp. 43–63. Sinauer,Sunderland, MA.

Hamrick JL, Godt MJW (1996) Effects of life history traits ongenetic diversity in plant species. Philosophical Transactions ofthe Royal Society of London, Series B, 351, 1291–1298.

Hamrick JL, Godt MJW, Murawski DA, Loveless MD (1991)Correlations between species traits and allozyme diversity:implications for conservation biology. In: Genetics and Con-servation of Rare Plants (eds Falk DA, Holsinger KE), pp. 75–86.Oxford University Press, New York.

Harry IB, Clarke DD (1987) The genetics of race-specific resist-ance in groundsel Senecio vulgaris L. to the powdery mildewfungus Erysiphe fischeri Blumer. New Phytologist, 107, 715–724.

Hermanutz LA, Weaver SE (1996) Agroecotypes or phenotypicplasticity? Comparison of agrestal and ruderal populations ofthe weed Solanum ptycanthum. Oecologia, 105, 271–280.

Holm L, Doll J, Holm E, Pancho J, Herberger J (1997) WorldWeeds: Natural Histories and Distribution. John Wiley & Sons,New York.

Holt JS, LeBaron HM (1990) Significance and distribution ofherbicide resistance. Weed Technology, 4, 141–149.

Isabel N, Beaulieu J, Thériault P, Bousquet J (1999) Direct evidencefor biased gene diversity estimates from dominant randomamplified polymorphic DNA (RAPD) fingerprints. MolecularEcology, 8, 477–483.

Kadereit JW (1984) The origin of Senecio vulgaris (Asteraceae). PlantSystematics and Evolution, 145, 135–153.

Kees H (1991) Zur Problematik der Triazinresistenz bei Samen-unkräutern im Mais und in gärtnerischen Kulturen. GesundePflanze, 7, 216–220.

Legendre P, Vaudor A (1991) The R Package for Multivariate DataAnalysis. Département de sciences biologiques, Université deMontréal, Montréal, Québec.

Lynch M, Milligan BG (1994) Analysis of population genetic struc-ture with RAPD markers. Molecular Ecology, 3, 91–99.

Mantel N (1967) The detection of disease clustering and generalizedregression approach. Cancer Research, 27, 209–220.

Moodie M, Finch RP, Marshall G (1997) Analysis of genetic vari-ation in wild mustard (Sinapis arvensis) using molecular markers.Weed Science, 45, 102–107.

Müller-Schärer H, Frantzen J (1996) An emerging system manage-ment approach for biological weed control in crops: Seneciovulgaris as a research model. Weed Research, 36, 483–491.

Müller-Schärer H, Rieger S (1998) Epidemic spread of the rustfungus Puccinia lagenophorae and its impact on the competitiveability of Senecio vulgaris in celeriac during early development.Biocontrol Science and Technology, 8, 59–72.

Müller-Schärer H, Scheepens PC, Greaves M (2000) Biologicalcontrol of weeds in European crops: recent achievements andfuture work. Weed Research, 40, 83–98.

Müller-Schärer H, Wyss GS (1994) Das Gemeine Kreuzkraut (Seneciovulgaris L.): Problemunkraut und Möglichkeiten der biologischenBekämpfung. Zeitschrift für Pflanzenkrankheiten und Pflanzenschutz,Sonderheft, XIV, 201–209.

Payne RW, Lane PW, Digby PGN et al. (1993) GENSTAT 5 ReferenceManual. Clarendon Press, Oxford.

Persson HA, Lundquist K, Nybom H (1998) RAPD analysis ofgenetic variation within and among populations of Turk’s-caplily (Lilium martagon L.). Hereditas, 128, 213–220.

Prober SM, Spindler LH, Brown AHD (1998) Conservation of thegrassy white box woodlands: effects of remnant populationsize on genetic diversity in the allotetraploid herb Microserislanceolata. Conservation Biology, 12, 1279–1290.

Roberts HA (1964) Emergence and longevity in cultivated soil ofseeds of some annual weeds. Weed Research, 4, 296–307.

Roberts HA, Feast PM (1972) Fate of seeds of some annual weedsin different depths of cultivated and undisturbed soil. WeedResearch, 12, 316–324.

Rogers SO, Bendich AJ (1988). Extraction of DNA from Plant Tissues(Plant Molecular Biology Manual). Kluwer, Belgium.

Schlichting CD (1986) The evolution of phenotypic plasticity inplants. Annual Review of Ecological Systematics, 17, 667–693.

Schmid B (1992) Phenotypic variation in plants. Evolutionary Trendsin Plants, 6, 45–60.

Schmid B (1994) Effects of genetic diversity in experimental standsof Solidago altissima — evidence for the potential role of pathogensas selective agents in plant populations. Journal of Ecology, 82,165–175.

Schmidt K, Jensen K (2000) Genetic structure and AFLP variationof remnant populations in the rare plant Pedicularis palustris(Scrophulariaceae) and its relation to population size and repro-ductive components. American Journal of Botany, 87, 678–689.

Silvertown JW, Lovett Doust J (1993). Introduction to Plant PopulationEcology. Blackwell Scientific Publications, Oxford.

Southwood TRE (1978). Ecological Methods, 2nd edn. Chapman &Hall, London.

Steinger T, Körner C, Schmid B (1996) Long-term persistence in achanging climate: DNA analysis suggests very old ages of clonesof alpine Carex curvula. Oecologia, 105, 94–99.

Stewart CN, Excoffier L (1996) Assessing population genetic struc-ture and variability with RAPD data: application to Vacciniummacrocarpon (American Cranberry). Journal of EvolutionaryBiology, 9, 153–171.

Stewart CN, Porter DM (1995) RAPD profiling in biologicalconservation: an application to estimating clonal variation in rareand endangered Iliamna in Virginia. Biological Conservation, 74,135–142.




Warwick SI (1990) Allozyme and life history varition in fivenorthwardly colonizing North American weed species. PlantSystematics and Evolution, 169, 41–54.

Warwick SI, Black LD (1993) Electrophoretic variation in triazine-resistant and -susceptible populations of the allogamous weedBrassica rapa. Weed Research, 33, 105–114.

Whitlock M, McCauley DE (1999) Indirect measures of gene flowand migration: Fst ≠ 1/ (4Nm + 1). Heredity, 82, 117–125.

Williams JGK, Kubelik AR, Livak KJ, Rafalski JA, Tingey SV (1990)DNA polymorphisms amplified by arbitrary primers are usefulas genetic markers. Nucleic Acids Research, 18, 6531–6535.

Wright S (1951) The general structure of populations. Annals ofEugenetics, 15, 323–354.

Wyss GS, Müller-Schärer H (1999) Infection process and resistancein the weed pathosystem Senecio vulgaris-Puccinia lagenophoraeand implications for biological control. Canadian Journal of Botany,77, 361–369.

Young A, Boyle T, Brown T (1998) The population genetic con-sequences of habitat fragmentation for plants. Trends in Ecologyand Evolution, 11, 413–418.

Heinz Müller-Schärer is associate professor of ecology. His maininterests are experimental plant population biology, plant–insect/pathogen interactions, population management and especiallyweed biocontrol. Markus Fischer is research associate. His maininterests are biodiversity and the population biology of rare, ofclonal and of invasive plants. This study is part of a link projectbetween a coordinated European Research project on ‘BiologicalControl of Weeds in Crops’ (COST-816) and an interdisciplinaryresearch programme on ‘Biodiversity and Sustainable Management’within the Swiss Priority Programme Environment of the SwissNational Science Foundation.


genetic structure of the annual weed senecio vulgaris in relation to

Documents